{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.0,
  "global_step": 497,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0,
      "learning_rate": 0.0,
      "loss": 2.3905,
      "step": 1
    },
    {
      "epoch": 0.0,
      "learning_rate": 5.119160496196309e-05,
      "loss": 2.0554,
      "step": 2
    },
    {
      "epoch": 0.01,
      "learning_rate": 8.113677421644257e-05,
      "loss": 2.0789,
      "step": 3
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00010238320992392618,
      "loss": 1.8802,
      "step": 4
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00011886322578355742,
      "loss": 1.8873,
      "step": 5
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00013232837917840569,
      "loss": 1.7897,
      "step": 6
    },
    {
      "epoch": 0.01,
      "learning_rate": 0.00014371300415799555,
      "loss": 1.736,
      "step": 7
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00015357481488588928,
      "loss": 1.6938,
      "step": 8
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00016227354843288515,
      "loss": 1.6686,
      "step": 9
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00017005483074552052,
      "loss": 1.7334,
      "step": 10
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00017709385681420508,
      "loss": 1.7926,
      "step": 11
    },
    {
      "epoch": 0.02,
      "learning_rate": 0.00018351998414036877,
      "loss": 1.7739,
      "step": 12
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00018943144823663685,
      "loss": 1.7048,
      "step": 13
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019490460911995866,
      "loss": 1.7128,
      "step": 14
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.00019999999999999998,
      "loss": 1.7366,
      "step": 15
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0002,
      "loss": 1.6946,
      "step": 16
    },
    {
      "epoch": 0.03,
      "learning_rate": 0.0002,
      "loss": 1.6957,
      "step": 17
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0002,
      "loss": 1.7196,
      "step": 18
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0002,
      "loss": 1.7363,
      "step": 19
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0002,
      "loss": 1.6447,
      "step": 20
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0002,
      "loss": 1.7061,
      "step": 21
    },
    {
      "epoch": 0.04,
      "learning_rate": 0.0002,
      "loss": 1.6932,
      "step": 22
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0002,
      "loss": 1.7471,
      "step": 23
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0002,
      "loss": 1.7047,
      "step": 24
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0002,
      "loss": 1.7012,
      "step": 25
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0002,
      "loss": 1.6516,
      "step": 26
    },
    {
      "epoch": 0.05,
      "learning_rate": 0.0002,
      "loss": 1.7336,
      "step": 27
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0002,
      "loss": 1.7234,
      "step": 28
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0002,
      "loss": 1.6453,
      "step": 29
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0002,
      "loss": 1.6176,
      "step": 30
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0002,
      "loss": 1.681,
      "step": 31
    },
    {
      "epoch": 0.06,
      "learning_rate": 0.0002,
      "loss": 1.7067,
      "step": 32
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0002,
      "loss": 1.7016,
      "step": 33
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0002,
      "loss": 1.6456,
      "step": 34
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0002,
      "loss": 1.6212,
      "step": 35
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0002,
      "loss": 1.5794,
      "step": 36
    },
    {
      "epoch": 0.07,
      "learning_rate": 0.0002,
      "loss": 1.7015,
      "step": 37
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0002,
      "loss": 1.7294,
      "step": 38
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0002,
      "loss": 1.7193,
      "step": 39
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0002,
      "loss": 1.6009,
      "step": 40
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0002,
      "loss": 1.5901,
      "step": 41
    },
    {
      "epoch": 0.08,
      "learning_rate": 0.0002,
      "loss": 1.6606,
      "step": 42
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0002,
      "loss": 1.6788,
      "step": 43
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0002,
      "loss": 1.6182,
      "step": 44
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0002,
      "loss": 1.6394,
      "step": 45
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0002,
      "loss": 1.6126,
      "step": 46
    },
    {
      "epoch": 0.09,
      "learning_rate": 0.0002,
      "loss": 1.6979,
      "step": 47
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0002,
      "loss": 1.7148,
      "step": 48
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0002,
      "loss": 1.6238,
      "step": 49
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0002,
      "loss": 1.6563,
      "step": 50
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0002,
      "loss": 1.6537,
      "step": 51
    },
    {
      "epoch": 0.1,
      "learning_rate": 0.0002,
      "loss": 1.6734,
      "step": 52
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0002,
      "loss": 1.655,
      "step": 53
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0002,
      "loss": 1.6857,
      "step": 54
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0002,
      "loss": 1.6753,
      "step": 55
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0002,
      "loss": 1.5768,
      "step": 56
    },
    {
      "epoch": 0.11,
      "learning_rate": 0.0002,
      "loss": 1.6318,
      "step": 57
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0002,
      "loss": 1.6508,
      "step": 58
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0002,
      "loss": 1.6852,
      "step": 59
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0002,
      "loss": 1.6452,
      "step": 60
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0002,
      "loss": 1.665,
      "step": 61
    },
    {
      "epoch": 0.12,
      "learning_rate": 0.0002,
      "loss": 1.5988,
      "step": 62
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0002,
      "loss": 1.609,
      "step": 63
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0002,
      "loss": 1.6471,
      "step": 64
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0002,
      "loss": 1.5924,
      "step": 65
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0002,
      "loss": 1.6239,
      "step": 66
    },
    {
      "epoch": 0.13,
      "learning_rate": 0.0002,
      "loss": 1.6684,
      "step": 67
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0002,
      "loss": 1.622,
      "step": 68
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0002,
      "loss": 1.648,
      "step": 69
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0002,
      "loss": 1.6071,
      "step": 70
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0002,
      "loss": 1.6175,
      "step": 71
    },
    {
      "epoch": 0.14,
      "learning_rate": 0.0002,
      "loss": 1.6612,
      "step": 72
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0002,
      "loss": 1.6523,
      "step": 73
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0002,
      "loss": 1.6282,
      "step": 74
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0002,
      "loss": 1.6118,
      "step": 75
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0002,
      "loss": 1.6734,
      "step": 76
    },
    {
      "epoch": 0.15,
      "learning_rate": 0.0002,
      "loss": 1.7171,
      "step": 77
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0002,
      "loss": 1.6933,
      "step": 78
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0002,
      "loss": 1.6644,
      "step": 79
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0002,
      "loss": 1.5887,
      "step": 80
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0002,
      "loss": 1.6366,
      "step": 81
    },
    {
      "epoch": 0.16,
      "learning_rate": 0.0002,
      "loss": 1.6485,
      "step": 82
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0002,
      "loss": 1.6397,
      "step": 83
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0002,
      "loss": 1.6234,
      "step": 84
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0002,
      "loss": 1.6982,
      "step": 85
    },
    {
      "epoch": 0.17,
      "learning_rate": 0.0002,
      "loss": 1.6833,
      "step": 86
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0002,
      "loss": 1.5772,
      "step": 87
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0002,
      "loss": 1.6249,
      "step": 88
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0002,
      "loss": 1.6587,
      "step": 89
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0002,
      "loss": 1.6305,
      "step": 90
    },
    {
      "epoch": 0.18,
      "learning_rate": 0.0002,
      "loss": 1.6185,
      "step": 91
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0002,
      "loss": 1.6246,
      "step": 92
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0002,
      "loss": 1.7224,
      "step": 93
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0002,
      "loss": 1.6597,
      "step": 94
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0002,
      "loss": 1.5627,
      "step": 95
    },
    {
      "epoch": 0.19,
      "learning_rate": 0.0002,
      "loss": 1.7173,
      "step": 96
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0002,
      "loss": 1.6461,
      "step": 97
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0002,
      "loss": 1.7234,
      "step": 98
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0002,
      "loss": 1.6376,
      "step": 99
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0002,
      "loss": 1.6216,
      "step": 100
    },
    {
      "epoch": 0.2,
      "learning_rate": 0.0002,
      "loss": 1.629,
      "step": 101
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0002,
      "loss": 1.6102,
      "step": 102
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0002,
      "loss": 1.6745,
      "step": 103
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0002,
      "loss": 1.6286,
      "step": 104
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0002,
      "loss": 1.6598,
      "step": 105
    },
    {
      "epoch": 0.21,
      "learning_rate": 0.0002,
      "loss": 1.6504,
      "step": 106
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0002,
      "loss": 1.5839,
      "step": 107
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0002,
      "loss": 1.6188,
      "step": 108
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0002,
      "loss": 1.5822,
      "step": 109
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0002,
      "loss": 1.6612,
      "step": 110
    },
    {
      "epoch": 0.22,
      "learning_rate": 0.0002,
      "loss": 1.6816,
      "step": 111
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0002,
      "loss": 1.6315,
      "step": 112
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0002,
      "loss": 1.6885,
      "step": 113
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0002,
      "loss": 1.6621,
      "step": 114
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0002,
      "loss": 1.57,
      "step": 115
    },
    {
      "epoch": 0.23,
      "learning_rate": 0.0002,
      "loss": 1.6848,
      "step": 116
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0002,
      "loss": 1.6398,
      "step": 117
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0002,
      "loss": 1.6485,
      "step": 118
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0002,
      "loss": 1.6395,
      "step": 119
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0002,
      "loss": 1.6469,
      "step": 120
    },
    {
      "epoch": 0.24,
      "learning_rate": 0.0002,
      "loss": 1.7036,
      "step": 121
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0002,
      "loss": 1.6051,
      "step": 122
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0002,
      "loss": 1.6186,
      "step": 123
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0002,
      "loss": 1.6133,
      "step": 124
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0002,
      "loss": 1.5975,
      "step": 125
    },
    {
      "epoch": 0.25,
      "learning_rate": 0.0002,
      "loss": 1.6332,
      "step": 126
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0002,
      "loss": 1.6541,
      "step": 127
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0002,
      "loss": 1.6024,
      "step": 128
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0002,
      "loss": 1.7152,
      "step": 129
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0002,
      "loss": 1.6086,
      "step": 130
    },
    {
      "epoch": 0.26,
      "learning_rate": 0.0002,
      "loss": 1.5337,
      "step": 131
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0002,
      "loss": 1.6136,
      "step": 132
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0002,
      "loss": 1.6653,
      "step": 133
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0002,
      "loss": 1.6817,
      "step": 134
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0002,
      "loss": 1.6385,
      "step": 135
    },
    {
      "epoch": 0.27,
      "learning_rate": 0.0002,
      "loss": 1.6456,
      "step": 136
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0002,
      "loss": 1.6359,
      "step": 137
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0002,
      "loss": 1.6771,
      "step": 138
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0002,
      "loss": 1.5712,
      "step": 139
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0002,
      "loss": 1.602,
      "step": 140
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0002,
      "loss": 1.6111,
      "step": 141
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0002,
      "loss": 1.5877,
      "step": 142
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0002,
      "loss": 1.6348,
      "step": 143
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0002,
      "loss": 1.7024,
      "step": 144
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0002,
      "loss": 1.6192,
      "step": 145
    },
    {
      "epoch": 0.29,
      "learning_rate": 0.0002,
      "loss": 1.5776,
      "step": 146
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0002,
      "loss": 1.6441,
      "step": 147
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0002,
      "loss": 1.6482,
      "step": 148
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0002,
      "loss": 1.6178,
      "step": 149
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0002,
      "loss": 1.6355,
      "step": 150
    },
    {
      "epoch": 0.3,
      "learning_rate": 0.0002,
      "loss": 1.5593,
      "step": 151
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0002,
      "loss": 1.5698,
      "step": 152
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0002,
      "loss": 1.5678,
      "step": 153
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0002,
      "loss": 1.62,
      "step": 154
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0002,
      "loss": 1.6039,
      "step": 155
    },
    {
      "epoch": 0.31,
      "learning_rate": 0.0002,
      "loss": 1.59,
      "step": 156
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0002,
      "loss": 1.6015,
      "step": 157
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0002,
      "loss": 1.6087,
      "step": 158
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0002,
      "loss": 1.6578,
      "step": 159
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0002,
      "loss": 1.6405,
      "step": 160
    },
    {
      "epoch": 0.32,
      "learning_rate": 0.0002,
      "loss": 1.6017,
      "step": 161
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0002,
      "loss": 1.6179,
      "step": 162
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0002,
      "loss": 1.6227,
      "step": 163
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0002,
      "loss": 1.6423,
      "step": 164
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0002,
      "loss": 1.6251,
      "step": 165
    },
    {
      "epoch": 0.33,
      "learning_rate": 0.0002,
      "loss": 1.5917,
      "step": 166
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0002,
      "loss": 1.6317,
      "step": 167
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0002,
      "loss": 1.6692,
      "step": 168
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0002,
      "loss": 1.605,
      "step": 169
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0002,
      "loss": 1.5232,
      "step": 170
    },
    {
      "epoch": 0.34,
      "learning_rate": 0.0002,
      "loss": 1.6234,
      "step": 171
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0002,
      "loss": 1.6455,
      "step": 172
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0002,
      "loss": 1.6034,
      "step": 173
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0002,
      "loss": 1.5841,
      "step": 174
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0002,
      "loss": 1.5337,
      "step": 175
    },
    {
      "epoch": 0.35,
      "learning_rate": 0.0002,
      "loss": 1.6562,
      "step": 176
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0002,
      "loss": 1.5882,
      "step": 177
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0002,
      "loss": 1.5694,
      "step": 178
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0002,
      "loss": 1.5397,
      "step": 179
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0002,
      "loss": 1.6489,
      "step": 180
    },
    {
      "epoch": 0.36,
      "learning_rate": 0.0002,
      "loss": 1.6329,
      "step": 181
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0002,
      "loss": 1.5919,
      "step": 182
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0002,
      "loss": 1.6217,
      "step": 183
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0002,
      "loss": 1.6385,
      "step": 184
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0002,
      "loss": 1.5572,
      "step": 185
    },
    {
      "epoch": 0.37,
      "learning_rate": 0.0002,
      "loss": 1.5961,
      "step": 186
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0002,
      "loss": 1.6038,
      "step": 187
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0002,
      "loss": 1.6287,
      "step": 188
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0002,
      "loss": 1.5813,
      "step": 189
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0002,
      "loss": 1.6572,
      "step": 190
    },
    {
      "epoch": 0.38,
      "learning_rate": 0.0002,
      "loss": 1.5287,
      "step": 191
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0002,
      "loss": 1.6176,
      "step": 192
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0002,
      "loss": 1.5759,
      "step": 193
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0002,
      "loss": 1.6038,
      "step": 194
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0002,
      "loss": 1.6411,
      "step": 195
    },
    {
      "epoch": 0.39,
      "learning_rate": 0.0002,
      "loss": 1.5893,
      "step": 196
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0002,
      "loss": 1.6322,
      "step": 197
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0002,
      "loss": 1.5794,
      "step": 198
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0002,
      "loss": 1.5791,
      "step": 199
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0002,
      "loss": 1.571,
      "step": 200
    },
    {
      "epoch": 0.4,
      "learning_rate": 0.0002,
      "loss": 1.605,
      "step": 201
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0002,
      "loss": 1.6436,
      "step": 202
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0002,
      "loss": 1.5692,
      "step": 203
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0002,
      "loss": 1.554,
      "step": 204
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0002,
      "loss": 1.5167,
      "step": 205
    },
    {
      "epoch": 0.41,
      "learning_rate": 0.0002,
      "loss": 1.5829,
      "step": 206
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0002,
      "loss": 1.5956,
      "step": 207
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0002,
      "loss": 1.6104,
      "step": 208
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0002,
      "loss": 1.6315,
      "step": 209
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0002,
      "loss": 1.5989,
      "step": 210
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0002,
      "loss": 1.6264,
      "step": 211
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0002,
      "loss": 1.6117,
      "step": 212
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0002,
      "loss": 1.5914,
      "step": 213
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0002,
      "loss": 1.635,
      "step": 214
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0002,
      "loss": 1.6579,
      "step": 215
    },
    {
      "epoch": 0.43,
      "learning_rate": 0.0002,
      "loss": 1.5833,
      "step": 216
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0002,
      "loss": 1.6144,
      "step": 217
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0002,
      "loss": 1.5891,
      "step": 218
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0002,
      "loss": 1.686,
      "step": 219
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0002,
      "loss": 1.5758,
      "step": 220
    },
    {
      "epoch": 0.44,
      "learning_rate": 0.0002,
      "loss": 1.5959,
      "step": 221
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0002,
      "loss": 1.5442,
      "step": 222
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0002,
      "loss": 1.6172,
      "step": 223
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0002,
      "loss": 1.5731,
      "step": 224
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0002,
      "loss": 1.5893,
      "step": 225
    },
    {
      "epoch": 0.45,
      "learning_rate": 0.0002,
      "loss": 1.5916,
      "step": 226
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0002,
      "loss": 1.5708,
      "step": 227
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0002,
      "loss": 1.6058,
      "step": 228
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0002,
      "loss": 1.6219,
      "step": 229
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0002,
      "loss": 1.6034,
      "step": 230
    },
    {
      "epoch": 0.46,
      "learning_rate": 0.0002,
      "loss": 1.5768,
      "step": 231
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0002,
      "loss": 1.6018,
      "step": 232
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0002,
      "loss": 1.6391,
      "step": 233
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0002,
      "loss": 1.5701,
      "step": 234
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0002,
      "loss": 1.6155,
      "step": 235
    },
    {
      "epoch": 0.47,
      "learning_rate": 0.0002,
      "loss": 1.5618,
      "step": 236
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0002,
      "loss": 1.6141,
      "step": 237
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0002,
      "loss": 1.593,
      "step": 238
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0002,
      "loss": 1.5951,
      "step": 239
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0002,
      "loss": 1.5351,
      "step": 240
    },
    {
      "epoch": 0.48,
      "learning_rate": 0.0002,
      "loss": 1.5303,
      "step": 241
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0002,
      "loss": 1.5775,
      "step": 242
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0002,
      "loss": 1.6633,
      "step": 243
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0002,
      "loss": 1.5798,
      "step": 244
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0002,
      "loss": 1.508,
      "step": 245
    },
    {
      "epoch": 0.49,
      "learning_rate": 0.0002,
      "loss": 1.6129,
      "step": 246
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0002,
      "loss": 1.5965,
      "step": 247
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0002,
      "loss": 1.6359,
      "step": 248
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0002,
      "loss": 1.5636,
      "step": 249
    },
    {
      "epoch": 0.5,
      "learning_rate": 0.0002,
      "loss": 1.6096,
      "step": 250
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0002,
      "loss": 1.6604,
      "step": 251
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0002,
      "loss": 1.636,
      "step": 252
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0002,
      "loss": 1.5935,
      "step": 253
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0002,
      "loss": 1.568,
      "step": 254
    },
    {
      "epoch": 0.51,
      "learning_rate": 0.0002,
      "loss": 1.592,
      "step": 255
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0002,
      "loss": 1.608,
      "step": 256
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0002,
      "loss": 1.5886,
      "step": 257
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0002,
      "loss": 1.5759,
      "step": 258
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0002,
      "loss": 1.5917,
      "step": 259
    },
    {
      "epoch": 0.52,
      "learning_rate": 0.0002,
      "loss": 1.5944,
      "step": 260
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0002,
      "loss": 1.5704,
      "step": 261
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0002,
      "loss": 1.6144,
      "step": 262
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0002,
      "loss": 1.6006,
      "step": 263
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0002,
      "loss": 1.5905,
      "step": 264
    },
    {
      "epoch": 0.53,
      "learning_rate": 0.0002,
      "loss": 1.5818,
      "step": 265
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0002,
      "loss": 1.5921,
      "step": 266
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0002,
      "loss": 1.621,
      "step": 267
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0002,
      "loss": 1.5944,
      "step": 268
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0002,
      "loss": 1.5793,
      "step": 269
    },
    {
      "epoch": 0.54,
      "learning_rate": 0.0002,
      "loss": 1.6598,
      "step": 270
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0002,
      "loss": 1.6511,
      "step": 271
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0002,
      "loss": 1.5654,
      "step": 272
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0002,
      "loss": 1.5664,
      "step": 273
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0002,
      "loss": 1.5366,
      "step": 274
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0002,
      "loss": 1.5561,
      "step": 275
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0002,
      "loss": 1.5952,
      "step": 276
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0002,
      "loss": 1.641,
      "step": 277
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0002,
      "loss": 1.6358,
      "step": 278
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0002,
      "loss": 1.549,
      "step": 279
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0002,
      "loss": 1.5649,
      "step": 280
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0002,
      "loss": 1.6332,
      "step": 281
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0002,
      "loss": 1.6376,
      "step": 282
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0002,
      "loss": 1.6463,
      "step": 283
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0002,
      "loss": 1.5739,
      "step": 284
    },
    {
      "epoch": 0.57,
      "learning_rate": 0.0002,
      "loss": 1.6165,
      "step": 285
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0002,
      "loss": 1.6247,
      "step": 286
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0002,
      "loss": 1.6713,
      "step": 287
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0002,
      "loss": 1.5769,
      "step": 288
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0002,
      "loss": 1.5357,
      "step": 289
    },
    {
      "epoch": 0.58,
      "learning_rate": 0.0002,
      "loss": 1.5967,
      "step": 290
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0002,
      "loss": 1.5385,
      "step": 291
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0002,
      "loss": 1.597,
      "step": 292
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0002,
      "loss": 1.6207,
      "step": 293
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0002,
      "loss": 1.6185,
      "step": 294
    },
    {
      "epoch": 0.59,
      "learning_rate": 0.0002,
      "loss": 1.6459,
      "step": 295
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0002,
      "loss": 1.5957,
      "step": 296
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0002,
      "loss": 1.529,
      "step": 297
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0002,
      "loss": 1.592,
      "step": 298
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0002,
      "loss": 1.6349,
      "step": 299
    },
    {
      "epoch": 0.6,
      "learning_rate": 0.0002,
      "loss": 1.5442,
      "step": 300
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0002,
      "loss": 1.5587,
      "step": 301
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0002,
      "loss": 1.6123,
      "step": 302
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0002,
      "loss": 1.5774,
      "step": 303
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0002,
      "loss": 1.5884,
      "step": 304
    },
    {
      "epoch": 0.61,
      "learning_rate": 0.0002,
      "loss": 1.5629,
      "step": 305
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0002,
      "loss": 1.6084,
      "step": 306
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0002,
      "loss": 1.5774,
      "step": 307
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0002,
      "loss": 1.5431,
      "step": 308
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0002,
      "loss": 1.5862,
      "step": 309
    },
    {
      "epoch": 0.62,
      "learning_rate": 0.0002,
      "loss": 1.6478,
      "step": 310
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0002,
      "loss": 1.6192,
      "step": 311
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0002,
      "loss": 1.558,
      "step": 312
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0002,
      "loss": 1.595,
      "step": 313
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0002,
      "loss": 1.6029,
      "step": 314
    },
    {
      "epoch": 0.63,
      "learning_rate": 0.0002,
      "loss": 1.5992,
      "step": 315
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0002,
      "loss": 1.5867,
      "step": 316
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0002,
      "loss": 1.5594,
      "step": 317
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0002,
      "loss": 1.5805,
      "step": 318
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0002,
      "loss": 1.5573,
      "step": 319
    },
    {
      "epoch": 0.64,
      "learning_rate": 0.0002,
      "loss": 1.6034,
      "step": 320
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0002,
      "loss": 1.5217,
      "step": 321
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0002,
      "loss": 1.5571,
      "step": 322
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0002,
      "loss": 1.574,
      "step": 323
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0002,
      "loss": 1.5856,
      "step": 324
    },
    {
      "epoch": 0.65,
      "learning_rate": 0.0002,
      "loss": 1.6237,
      "step": 325
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0002,
      "loss": 1.6045,
      "step": 326
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0002,
      "loss": 1.5899,
      "step": 327
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0002,
      "loss": 1.5986,
      "step": 328
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0002,
      "loss": 1.5444,
      "step": 329
    },
    {
      "epoch": 0.66,
      "learning_rate": 0.0002,
      "loss": 1.6621,
      "step": 330
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0002,
      "loss": 1.5676,
      "step": 331
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0002,
      "loss": 1.6445,
      "step": 332
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0002,
      "loss": 1.6251,
      "step": 333
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0002,
      "loss": 1.5704,
      "step": 334
    },
    {
      "epoch": 0.67,
      "learning_rate": 0.0002,
      "loss": 1.5855,
      "step": 335
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0002,
      "loss": 1.5764,
      "step": 336
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0002,
      "loss": 1.595,
      "step": 337
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0002,
      "loss": 1.6169,
      "step": 338
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0002,
      "loss": 1.5623,
      "step": 339
    },
    {
      "epoch": 0.68,
      "learning_rate": 0.0002,
      "loss": 1.5863,
      "step": 340
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0002,
      "loss": 1.5875,
      "step": 341
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0002,
      "loss": 1.5778,
      "step": 342
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0002,
      "loss": 1.5129,
      "step": 343
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0002,
      "loss": 1.5755,
      "step": 344
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0002,
      "loss": 1.5552,
      "step": 345
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002,
      "loss": 1.54,
      "step": 346
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002,
      "loss": 1.5649,
      "step": 347
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002,
      "loss": 1.6039,
      "step": 348
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002,
      "loss": 1.6116,
      "step": 349
    },
    {
      "epoch": 0.7,
      "learning_rate": 0.0002,
      "loss": 1.6058,
      "step": 350
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002,
      "loss": 1.5868,
      "step": 351
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002,
      "loss": 1.5545,
      "step": 352
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002,
      "loss": 1.5983,
      "step": 353
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002,
      "loss": 1.5173,
      "step": 354
    },
    {
      "epoch": 0.71,
      "learning_rate": 0.0002,
      "loss": 1.5648,
      "step": 355
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002,
      "loss": 1.5758,
      "step": 356
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002,
      "loss": 1.4697,
      "step": 357
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002,
      "loss": 1.5784,
      "step": 358
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002,
      "loss": 1.5836,
      "step": 359
    },
    {
      "epoch": 0.72,
      "learning_rate": 0.0002,
      "loss": 1.5322,
      "step": 360
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002,
      "loss": 1.5725,
      "step": 361
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002,
      "loss": 1.5975,
      "step": 362
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002,
      "loss": 1.6772,
      "step": 363
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002,
      "loss": 1.5686,
      "step": 364
    },
    {
      "epoch": 0.73,
      "learning_rate": 0.0002,
      "loss": 1.6235,
      "step": 365
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002,
      "loss": 1.665,
      "step": 366
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002,
      "loss": 1.6541,
      "step": 367
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002,
      "loss": 1.6217,
      "step": 368
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002,
      "loss": 1.5859,
      "step": 369
    },
    {
      "epoch": 0.74,
      "learning_rate": 0.0002,
      "loss": 1.5372,
      "step": 370
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002,
      "loss": 1.6096,
      "step": 371
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002,
      "loss": 1.608,
      "step": 372
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002,
      "loss": 1.5501,
      "step": 373
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002,
      "loss": 1.5739,
      "step": 374
    },
    {
      "epoch": 0.75,
      "learning_rate": 0.0002,
      "loss": 1.5095,
      "step": 375
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002,
      "loss": 1.6264,
      "step": 376
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002,
      "loss": 1.5659,
      "step": 377
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002,
      "loss": 1.6025,
      "step": 378
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002,
      "loss": 1.5658,
      "step": 379
    },
    {
      "epoch": 0.76,
      "learning_rate": 0.0002,
      "loss": 1.5672,
      "step": 380
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002,
      "loss": 1.5334,
      "step": 381
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002,
      "loss": 1.6729,
      "step": 382
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002,
      "loss": 1.5454,
      "step": 383
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002,
      "loss": 1.6536,
      "step": 384
    },
    {
      "epoch": 0.77,
      "learning_rate": 0.0002,
      "loss": 1.5829,
      "step": 385
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002,
      "loss": 1.5635,
      "step": 386
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002,
      "loss": 1.6348,
      "step": 387
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002,
      "loss": 1.5718,
      "step": 388
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002,
      "loss": 1.6004,
      "step": 389
    },
    {
      "epoch": 0.78,
      "learning_rate": 0.0002,
      "loss": 1.5801,
      "step": 390
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002,
      "loss": 1.612,
      "step": 391
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002,
      "loss": 1.5958,
      "step": 392
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002,
      "loss": 1.5776,
      "step": 393
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002,
      "loss": 1.5179,
      "step": 394
    },
    {
      "epoch": 0.79,
      "learning_rate": 0.0002,
      "loss": 1.5704,
      "step": 395
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0002,
      "loss": 1.5666,
      "step": 396
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0002,
      "loss": 1.6144,
      "step": 397
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0002,
      "loss": 1.5494,
      "step": 398
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0002,
      "loss": 1.5865,
      "step": 399
    },
    {
      "epoch": 0.8,
      "learning_rate": 0.0002,
      "loss": 1.5612,
      "step": 400
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0002,
      "loss": 1.5335,
      "step": 401
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0002,
      "loss": 1.5394,
      "step": 402
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0002,
      "loss": 1.654,
      "step": 403
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0002,
      "loss": 1.5618,
      "step": 404
    },
    {
      "epoch": 0.81,
      "learning_rate": 0.0002,
      "loss": 1.5648,
      "step": 405
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0002,
      "loss": 1.5948,
      "step": 406
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0002,
      "loss": 1.5983,
      "step": 407
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0002,
      "loss": 1.5884,
      "step": 408
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0002,
      "loss": 1.5481,
      "step": 409
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0002,
      "loss": 1.5742,
      "step": 410
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0002,
      "loss": 1.5628,
      "step": 411
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0002,
      "loss": 1.538,
      "step": 412
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0002,
      "loss": 1.6311,
      "step": 413
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0002,
      "loss": 1.5706,
      "step": 414
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0002,
      "loss": 1.6155,
      "step": 415
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0002,
      "loss": 1.6288,
      "step": 416
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0002,
      "loss": 1.5604,
      "step": 417
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0002,
      "loss": 1.6095,
      "step": 418
    },
    {
      "epoch": 0.84,
      "learning_rate": 0.0002,
      "loss": 1.5654,
      "step": 419
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0002,
      "loss": 1.535,
      "step": 420
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0002,
      "loss": 1.549,
      "step": 421
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0002,
      "loss": 1.489,
      "step": 422
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0002,
      "loss": 1.5764,
      "step": 423
    },
    {
      "epoch": 0.85,
      "learning_rate": 0.0002,
      "loss": 1.5929,
      "step": 424
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0002,
      "loss": 1.5851,
      "step": 425
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0002,
      "loss": 1.5813,
      "step": 426
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0002,
      "loss": 1.5122,
      "step": 427
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0002,
      "loss": 1.5548,
      "step": 428
    },
    {
      "epoch": 0.86,
      "learning_rate": 0.0002,
      "loss": 1.5466,
      "step": 429
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0002,
      "loss": 1.583,
      "step": 430
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0002,
      "loss": 1.5963,
      "step": 431
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0002,
      "loss": 1.5596,
      "step": 432
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0002,
      "loss": 1.5567,
      "step": 433
    },
    {
      "epoch": 0.87,
      "learning_rate": 0.0002,
      "loss": 1.6284,
      "step": 434
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0002,
      "loss": 1.4996,
      "step": 435
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0002,
      "loss": 1.6172,
      "step": 436
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0002,
      "loss": 1.6038,
      "step": 437
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0002,
      "loss": 1.576,
      "step": 438
    },
    {
      "epoch": 0.88,
      "learning_rate": 0.0002,
      "loss": 1.5971,
      "step": 439
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0002,
      "loss": 1.5976,
      "step": 440
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0002,
      "loss": 1.5626,
      "step": 441
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0002,
      "loss": 1.6477,
      "step": 442
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0002,
      "loss": 1.5973,
      "step": 443
    },
    {
      "epoch": 0.89,
      "learning_rate": 0.0002,
      "loss": 1.588,
      "step": 444
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0002,
      "loss": 1.5463,
      "step": 445
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0002,
      "loss": 1.5769,
      "step": 446
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0002,
      "loss": 1.59,
      "step": 447
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0002,
      "loss": 1.6125,
      "step": 448
    },
    {
      "epoch": 0.9,
      "learning_rate": 0.0002,
      "loss": 1.5225,
      "step": 449
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0002,
      "loss": 1.5926,
      "step": 450
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0002,
      "loss": 1.5467,
      "step": 451
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0002,
      "loss": 1.5346,
      "step": 452
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0002,
      "loss": 1.5915,
      "step": 453
    },
    {
      "epoch": 0.91,
      "learning_rate": 0.0002,
      "loss": 1.5483,
      "step": 454
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0002,
      "loss": 1.6008,
      "step": 455
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0002,
      "loss": 1.5863,
      "step": 456
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0002,
      "loss": 1.5734,
      "step": 457
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0002,
      "loss": 1.4915,
      "step": 458
    },
    {
      "epoch": 0.92,
      "learning_rate": 0.0002,
      "loss": 1.536,
      "step": 459
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0002,
      "loss": 1.5833,
      "step": 460
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0002,
      "loss": 1.5531,
      "step": 461
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0002,
      "loss": 1.5145,
      "step": 462
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0002,
      "loss": 1.5802,
      "step": 463
    },
    {
      "epoch": 0.93,
      "learning_rate": 0.0002,
      "loss": 1.5289,
      "step": 464
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0002,
      "loss": 1.5796,
      "step": 465
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0002,
      "loss": 1.6064,
      "step": 466
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0002,
      "loss": 1.5794,
      "step": 467
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0002,
      "loss": 1.5762,
      "step": 468
    },
    {
      "epoch": 0.94,
      "learning_rate": 0.0002,
      "loss": 1.5946,
      "step": 469
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0002,
      "loss": 1.5842,
      "step": 470
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0002,
      "loss": 1.5742,
      "step": 471
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0002,
      "loss": 1.5489,
      "step": 472
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0002,
      "loss": 1.5838,
      "step": 473
    },
    {
      "epoch": 0.95,
      "learning_rate": 0.0002,
      "loss": 1.5397,
      "step": 474
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0002,
      "loss": 1.5383,
      "step": 475
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0002,
      "loss": 1.5586,
      "step": 476
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0002,
      "loss": 1.5401,
      "step": 477
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0002,
      "loss": 1.6091,
      "step": 478
    },
    {
      "epoch": 0.96,
      "learning_rate": 0.0002,
      "loss": 1.5577,
      "step": 479
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0002,
      "loss": 1.5065,
      "step": 480
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0002,
      "loss": 1.5649,
      "step": 481
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0002,
      "loss": 1.5487,
      "step": 482
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0002,
      "loss": 1.5234,
      "step": 483
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0002,
      "loss": 1.6588,
      "step": 484
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0002,
      "loss": 1.5487,
      "step": 485
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0002,
      "loss": 1.6292,
      "step": 486
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0002,
      "loss": 1.4982,
      "step": 487
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0002,
      "loss": 1.5354,
      "step": 488
    },
    {
      "epoch": 0.98,
      "learning_rate": 0.0002,
      "loss": 1.5821,
      "step": 489
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0002,
      "loss": 1.5866,
      "step": 490
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0002,
      "loss": 1.6163,
      "step": 491
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0002,
      "loss": 1.5639,
      "step": 492
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0002,
      "loss": 1.6021,
      "step": 493
    },
    {
      "epoch": 0.99,
      "learning_rate": 0.0002,
      "loss": 1.5625,
      "step": 494
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0002,
      "loss": 1.5594,
      "step": 495
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0002,
      "loss": 1.5592,
      "step": 496
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0002,
      "loss": 1.4951,
      "step": 497
    },
    {
      "epoch": 1.0,
      "step": 497,
      "total_flos": 46997707161600.0,
      "train_loss": 1.6110975456909395,
      "train_runtime": 13532.02,
      "train_samples_per_second": 9.395,
      "train_steps_per_second": 0.037
    }
  ],
  "max_steps": 497,
  "num_train_epochs": 1,
  "total_flos": 46997707161600.0,
  "trial_name": null,
  "trial_params": null
}